DOM105 2019

Session 1
Reading: SfM Ch.2,3
Categorical and Numerical Data
Categorical data is data that is separated
into various groupings or categories for
display.
Takes form of tables, bar charts, pie
charts, etc.
Numerical data comprises of numbers that
have not been separated into categories.
Displays of numerical data include arrays,
frequency distributions, scatter plots, etc.
Both types of data can be displayed using
some types of tables such as Pivot Tables.
Summary table
Tallies the values of various categories as frequencies and
percentages for each category.
Contingency Table
Cross-tabulates, or tallies jointly, the value
of two or more categorical variables,
allowing the study of patterns. Tallies of
frequency, or percentages.
Display Categorical Data – Bar Chart

Investor's Portfolio

Savings
CD

Bonds
Stocks

0 10 20 30 40 50
Amount in K$
Pie Chart – Investor’s portfolio
Savings
15%
Stocks
42%
CD
14%

Percentages are rounded to

Bonds the nearest percent.
29%
Side by side Bar Chart
Uses sets of bars to show joint responses from two or more
categorical variables.

Comparing Investors

Savings

CD

Bonds

Stocks

0 10 20 30 40 50 60

Investor A Investor B Investor C

Tabulating Numerical Data: Frequency Distributions

Sort raw data in ascending order:

12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58
Find range: 58 - 12 = 46
Select number of classes: 5 (usually between 5 and 15)
Compute class interval (width): 10 (46/5 then round up)
Determine class boundaries (limits): 10, 20, 30, 40, 50, 60
Compute class midpoints: 15, 25, 35, 45, 55
Count observations & assign to classes
Frequency Distributions, Relative Frequency
Distributions and Percentage Distributions

Data in ordered array:

12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58

Relative
Class Frequency Frequency Percentage
10 but under 20 3 .15 15
20 but under 30 6 .30 30
30 but under 40 5 .25 25
40 but under 50 4 .20 20
50 but under 60 2 .10 10
Total 20 1 100
© 2002 Prentice-Hall, Inc.
Graphing Numerical Data:
The Histogram

Data in ordered array:

12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58
Histogram

7 6
Frequency 6 5
5 4 No Gaps
4 3 Between
3 2 Bars
2
1 0 0
0
5 15 25 36 45 55 More

Class Boundaries
Class Midpoints
Tabulating Numerical Data:
Cumulative Frequency

Data in ordered array:

12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58

Cumulative Cumulative
Class Frequency % Frequency
10 but under 20 3 15
20 but under 30 9 45
30 but under 40 14 70
40 but under 50 18 90
50 but under 60 20 100
Graphing Numerical Data:
The Ogive (Cumulative % Polygon)

Data in ordered array:

12, 13, 17, 21, 24, 24, 26, 27, 27, 30, 32, 35, 37, 38, 41, 43, 44, 46, 53, 58

Ogive

100
80
60
40
20
0
10 20 30 40 50 60

Class Boundaries (Not Midpoints)

Graphing Two Numerical Variables - Scatter Plot

Mutual Funds Scatter Plot

40
Total Year to
Date Return
(%) 30
20
10
0
0 10 20 30 40
Net Asset Values
Time-Series Plot
Numerical variable on Y-axis, associated time period on X-
axis.
Errors in visualizing data
Using “chart junk”, visual effects that distort
or distract from the data to be presented, eg:
garish graphics, irrelevant visuals.
Failing to provide a relative basis in
comparing data between groups. For
example, two separate pie charts showing the
operations of two companies does not help if
we’re trying to compare the two.
Compressing the vertical axis – using an axis
going up to 100 when the highest value is 30.
Make sure to include zero on the axes.
Measures of Central Tendency
Most sets of data show a tendency to
group around a central value, this is the
‘central tendency’.
The most common measures of central
tendency are mean, median, and mode.
The mean, also called the arithmetic
mean, is the average of all values in the
sample space.
n

X i
X1  X 2    X n
X i 1

n n
n is the size of the sample.
Median

 Robust measure of central tendency

 Not affected by extreme values
 In an ordered array, the median is
the “middle” number
 Median: (n+1)/2 ranked value.
 If n is odd, the median is the middle
number.
 If n is even, the median is the
average of the two middle numbers.
Mode
A measure of central tendency
It is the value that occurs most often
in the sample.
Not affected by extreme values
Used for either numerical or
categorical data
There may be no mode if all values
have the same frequency
There may be several modes if more
than one value are tied for the
highest frequency.
Variation and shape of data: Range
 Measure of variation
 Difference between the largest and the
Range
smallest  X Largest  X Smallest
observations:

 Ignores the way in which data are distributed

 Does not consider how the values cluster

between extremes.
Quartiles
 
Quartiles split data into 4 parts.
1st Quartile splits the lowest 25% of the values from the rest.

3rd Quartile splits the lowest 75% of the values from the rest.

Q2 is the median.

If the rank is a half (2.5th, 7.5th etc.) then the quartile is
average of the two values on either side. If the rank is a
fraction other than half, round to nearest integer.
Interquartile Range
Measure of variation
Also known as midspread
Spread in the middle 50%
Difference between the first and third quartiles

Not affected by extreme values

Data in Ordered Array: 11 12 13 16 16 17 17 18 21

Interquartile Range  Q3  Q1  17.5  12.5  5

Percentile
To find top xth percentile, we use same method as quartile.
List data in ascending order
xth percentile = Data in rank (n+1)x/100, where n is
number of data points.
In case of fractional value of rank, use unitary method to
find value.
Eg: 80th percentile out of 30 data points would be
31*0.8=24.8th rank.
Value would be 24th data point * 0.2 + 25th data point * 0.8
Variance
Important

  measure of variation
Shows variation about the mean
Is the average of the square of the difference between an element and the
mean
n

 X X
2
i
S2  i 1
Sample variance: n 1 , is the sample mean.
N

 Xi   
2

2  i 1

N
Population variance: , µ is the population
mean.
Standard Deviation
Most important measure of variation
Shows variation about the mean
Has the same units as the original data
Is the square root of the variance
n

 X X
2
Sample standard deviation: i
S i 1

n 1
N

 Xi   
2
Population standard deviation:
 i 1

N
Why do we divide by (n-1) for sample variance?
 For a sample variance to be unbiased, the average variance for
all possible samples for a given population has to be equal to
the population variance.
 It was mathematically shown that if the sample variance was
calculated using n instead of n-1, the average variance of all
possible samples was not equal to population variance.
 This is called Bessel’s correction.
 Only used when the population mean and variance is unknown.
Shape of a Distribution
Describes how data is distributed
Measures of shape
Symmetric or skewed

Left-skewed or negative Right-skewed or positive

Symmetric
Mean < Median < Mode Mean = Median =Mode Mode < Median < Mean
5-number summary and the Box Plot
The 5 numbers: smallest X, Q , Q , Q , largest X
1 2 3
Boxplot
Graphical display of data using 5-number summary

Median( Q2 ) Xlargest
X smallest Q Q3
1

4 6 8 10 12
Distribution Shape and the Boxplot

Left-Skewed Symmetric Right-Skewed

Q1 Q2 Q3 Q1 Q2Q3 Q1 Q2 Q3
Measuring skewness
 Skewness is the measure of asymmetry in a data
distribution. One method of calculating it is adjusted Fisher
Pearson coefficient, as follows:

A symmetrical distribution like a normal distribution will

have . Negative value indicates left-skewed data, positive
indicates right-skewed.
Presence of extreme outliers can distort value of G, giving
erroneous results.
Measuring kurtosis
Kurtosis is a measure of how ‘heavy’ the tails of a data set
are, i.e., how many outliers are present, relative to a
normal distribution.

Normal distribution has kurtosis = 0. Higher kurtosis

indicates large number of outliers, lower means few
outliers.
Like skewness, extreme outliers can distort kurtosis values.
Pitfalls and Ethical Considerations
Data analysis is objective
Should report the summary measures that best meet the
assumptions about the data set
Data interpretation is subjective
Should be done in fair, neutral and clear manner

Numerical descriptive measures:

Should document both good and bad results
Should be presented in a fair, objective and neutral manner
Should not use inappropriate summary measures to distort facts